gopherenum-breadth.c

harvest是一个下载html网页得机器人

    int y;

    if ((tup = url_open(url)) == NULL)
	return 0;;
    if (url_in_db(tup->url)) {	/* Have we been here? */
	Debug(43, 1, ("Already Visited URL: %s\n", tup->url));
	url_close(tup);
	return 0;
    }
    if ((y = filter_selection(tup))) {
	Debug(43, 1, ("Removing Candidate: [%s] %s\n",
		Filter_Type_Name[y], tup->url));
	if (not_visited)
	    fprintf(not_visited, "[FILTER] %s\n", tup->url);
	url_close(tup);
	return 0;
    }
    if (!visit_server(tup)) {
	Debug(43, 1, ("Server count exceeded: %s\n",
		tup->url));
	if (not_visited)
	    fprintf(not_visited, "[SERVER] %s\n", tup->url);
	url_close(tup);
	return 0;
    }
    if (!RobotsTxtCheck(tup)) {
	Debug(43, 1, ("Disallowed by robots.txt: %s\n", tup->url));
	if (not_visited)
	    fprintf(not_visited, "[ROBOTS.TXT] %s\n", tup->url);
	url_close(tup);
	return 0;
    }
    return 1;
}

static int gopher_enum(up, depth)
     URL *up;
     int depth;
{
    FILE *fp = NULL;
    char *s = NULL;
    char *p = NULL;
    char *q = NULL;
    char *gopher_name = NULL;
    char *gopher_path = NULL;
    char *gopher_host = NULL;
    char *gopher_port = NULL;
    int nurls = 0;
    static char buf[BUFSIZ];
    static char urlbuf[BUFSIZ];
    static char newurl[BUFSIZ];

    if (url_in_db(up->url)) {	/* Have we been here? */
	Debug(43, 1, ("Already Visited URL: %s\n", up->url));
	return 0;
    }
    if (url_retrieve(up)) {	/* Grab the URL; success? */
	Debug(43, 1, ("Cannot Retrieve URL: %s\n", up->url));

#ifdef DONT_RETRY_FAILS
        mark_failed(up);
#endif

	return 0;
    }
    if (up->md5 && md5_in_db(up->md5)) {	/* Have we been here? */
	Debug(43, 1, ("Already Visited MD5: %s\n", up->url));
	return 0;
    }
    /* Remember that we've been here before */
    if (up->md5)
	mark_retrieved(up);
    if (up->gophertype == 0)
	return 0;

    /*
     *  For each meny entry, convert it to a URL, and add it to
     *  the global list of URLs to process.
     */
    if ((fp = fopen(up->filename, "r")) == NULL) {
	log_errno2(__FILE__, __LINE__, up->filename);
	return 0;
    }
    while (fgets(buf, BUFSIZ, fp)) {
	if ((s = strchr(buf, '\r')))
	    *s = (char) '\n';
	strcpy(urlbuf, buf);
	if ((s = strchr(buf, '\n')))
	    *s = (char) '\0';
	Debug(43, 5, ("Input: %s\n", buf));
	if (!strcmp(buf, "."))
	    break;

	p = urlbuf;
	if ((q = strchr(p, '\t')) == NULL) {
	    errorlog("Illegal Gopher format: No Name: %s\n", buf);
	    goto gopher_enum_cont;
	}
	*q = (char) '\0';
	gopher_name = xstrdup(p);
	Debug(43, 5, ("gopher_name = '%s'\n", gopher_name));

	p = q + 1;
	if ((q = strchr(p, '\t')) == NULL) {
	    errorlog("Illegal Gopher format: No Path: %s\n", buf);
	    goto gopher_enum_cont;
	}
	*q = (char) '\0';
	gopher_path = xstrdup(rfc1738_escape(p));
	Debug(43, 5, ("gopher_path = '%s'\n", gopher_path));

	p = q + 1;
	if ((q = strchr(p, '\t')) == NULL) {
	    errorlog("Illegal Gopher format: No Host: %s\n", buf);
	    goto gopher_enum_cont;
	}
	*q = (char) '\0';
	gopher_host = xstrdup(p);
	Debug(43, 5, ("gopher_host = '%s'\n", gopher_host));

	p = q + 1;
	q = strchr(p, '\t');
	if (q == NULL)
	    q = strchr(p, '\n');
	if (q == NULL) {
	    errorlog("Illegal Gopher format: No Port: %s\n", buf);
	    goto gopher_enum_cont;
	}
	*q = (char) '\0';
	gopher_port = xstrdup(p);
	Debug(43, 5, ("gopher_port = '%s'\n", gopher_port));

	/* Fix for wierd cross-site Gopher links - wessels */
	if (!strncasecmp(gopher_path, "ftp%3a", 6))
	    goto gopher_enum_cont;
	if (!strncasecmp(gopher_path, "ftp:", 4))
	    goto gopher_enum_cont;
	if (!strncasecmp(gopher_path, "exec%3a", 7))
	    goto gopher_enum_cont;
	if (!strncasecmp(gopher_path, "exec:", 5))
	    goto gopher_enum_cont;

	sprintf(newurl, "gopher://%s:%d/%c%s", gopher_host,
	    atoi(gopher_port), gopher_name[0], gopher_path);
	if (url_is_allowed(newurl)) {
	    add_to_list(newurl, depth);
	    nurls++;
	}

      gopher_enum_cont:
	xfree(gopher_name);
	gopher_name = NULL;
	xfree(gopher_path);
	gopher_path = NULL;
	xfree(gopher_host);
	gopher_host = NULL;
	xfree(gopher_port);
	gopher_port = NULL;

    }
    fclose(fp);
    Debug(43, 1, ("Adding %d URLs from %s to workload\n", nurls, up->url));
    return 1;
}

/* ---------------------------------------------------------------------- */

/*
 *  initialize() - Basic init routines
 */
static void initialize()
{
    char *s = NULL;
    extern int liburl_conform_rfc1738;
    FILE *logfp = NULL;

#ifdef USE_HOST_CACHE
    host_cache_init();
#endif

    cur_depth = max_depth = url_max = host_max = 0;
    if ((s = getenv("HARVEST_URL_MAX")) != NULL)
	url_max = atoi(s);
    if ((s = getenv("HARVEST_HOST_MAX")) != NULL)
	host_max = atoi(s);
    if ((s = getenv("HARVEST_DEPTH_MAX")) != NULL)
	max_depth = atoi(s);
    if ((s = getenv("HARVEST_DEPTH_CUR")) != NULL)
	cur_depth = atoi(s);
    if (url_max < 1)
	url_max = 250;		/* hard-coded maximum */
    if (host_max < 1)
	host_max = 1;		/* hard-coded maximum */
    if (max_depth < 1)
	max_depth = 0;		/* hard-coded maximum */
    host_filterfile = getenv("HARVEST_HOST_FILTER");
    url_filterfile = getenv("HARVEST_URL_FILTER");
    access_types = getenv("HARVEST_ACCESS_TYPES");

    if ((s = getenv("HARVEST_GATHERER_LOGFILE")) != (char *) NULL)
	logfp = fopen(s, "a+");
    if (logfp == (FILE *) NULL)
	logfp = stderr;
    init_log3("gopherenum-breadth", logfp, stderr);
    init_url();
    liburl_conform_rfc1738 = 1;
    filter_initialize();
    Debug(43, 5, ("access_mask: %#02X\n", access_mask));

    /* Open GDBM databases to keep track of where we've been */
    urldb_filename = xstrdup(tempnam(NULL, "Gurl"));
    urldbf = gdbm_open(urldb_filename, 0, GDBM_NEWDB, 0644, NULL);
    if (urldbf == NULL) {
	log_errno(urldb_filename);
	fatal("gdbm_open: %s: %s", urldb_filename,
	    gdbm_strerror(gdbm_errno));
    }
    hostdb_filename = xstrdup(tempnam(NULL, "Ghost"));
    hostdbf = gdbm_open(hostdb_filename, 0, GDBM_NEWDB, 0644, NULL);
    if (hostdbf == NULL) {
	log_errno(hostdb_filename);
	fatal("gdbm_open: %s: %s", hostdb_filename,
	    gdbm_strerror(gdbm_errno));
    }
    md5db_filename = xstrdup(tempnam(NULL, "Gmd5"));
    md5dbf = gdbm_open(md5db_filename, 0, GDBM_NEWDB, 0644, NULL);
    if (md5dbf == NULL) {
	log_errno(md5db_filename);
	fatal("gdbm_open: %s: %s", md5db_filename,
	    gdbm_strerror(gdbm_errno));
    }
    /* open not-visited file */
    if ((s = getenv("HARVEST_NOT_VISITED_LOG")) != NULL)
	not_visited = fopen(s, "a+");
    if (not_visited)
	setbuf(not_visited, NULL);
}

/* Die gracefully */
static void sigdie(x)
     int x;
{
    int i;
#ifdef USE_HOST_CACHE
    dump_host_cache(43, 9);
#endif
    if (urldbf != NULL)
	gdbm_close(urldbf);
    if (hostdbf != NULL)
	gdbm_close(hostdbf);
    if (md5dbf != NULL)
	gdbm_close(md5dbf);
    if (not_visited)
	fclose(not_visited);

    /* (void) unlink(urldb_filename); */
    crremove(urldb_filename);
    xfree(urldb_filename);
    /* (void) unlink(hostdb_filename); */
    crremove(hostdb_filename);
    xfree(hostdb_filename);
    /* (void) unlink(md5db_filename); */
    crremove(md5db_filename);
    xfree(md5db_filename);

    for (i = 0; i < 100; i++) {
	if (i > max_depth && depth_hist[i] == 0)
	    break;
	Log("Found %8d objects at depth %d\n", depth_hist[i], i);
    }

    Debug(43, 1, ("gopherenum-breadth: exiting (signal %d)\n", x));
    exit(0);
}

/* ---------------------------------------------------------------------- */

static void usage()
{
    fprintf(stderr, "Usage: gopherenum-breadth gopher-URL\n");
    exit(1);
}

int main(argc, argv)
     int argc;
     char **argv;
{
    URL *up = NULL;
    list_t *l = NULL;
    char *url = NULL;
    int depth = 0;

    debug_init();		/* from $HARVEST_DEBUG */
    for (argc--, argv++; argc > 0 && **argv == '-'; argc--, argv++) {
	if (!strncmp(*argv, "-D", 2)) {
	    debug_flag(*argv);
	}
    }

    if (argc != 1)
	usage();

    for (depth = 0; depth < 100; depth++)
	depth_hist[depth] = 0;

    signal(SIGTERM, sigdie);	/* Die gracefully */
    signal(SIGINT, sigdie);
    signal(SIGPIPE, sigdie);	/* Quickly clean up on broken pipe */

    initialize();		/* Initialize */

    Debug(43, 1, ("gopherenum-breadth: Starting...\n"));

    /* Grab the RootNode URL from the command line */
    if ((up = url_open(*argv)) == NULL || up->type != URL_GOPHER) {
	usage();
    }
    /* Mark the RootNode */
    tree_root = xstrdup(up->url);
    Tail = &head;

    /*
     * [e-mail] -- Gatherer visits too many hosts
     * 6/3/96. Make sure the first URL we start with is added
     *         to the list of servers visited.
     */
    (void)visit_server(up);
    printf("%s\n", up->url);	/* Print tree root */
    add_to_list(up->url, cur_depth);	/* start at depth = 0 */
    url_close(up);

    for (l = head; l; l = free_from_list(l)) {
	url = (char *) l->ptr;
	depth = l->depth;

	if (depth < 100)
	    depth_hist[depth]++;

	if (max_depth > 0 && depth > max_depth) {
	    if (not_visited)
		fprintf(not_visited, "[DEPTH] %s\n", url);
	    Debug(43, 1, ("Maximum Depth of %d Reached: %s\n",
		    max_depth, url));
	    continue;
	}
	Debug(43, 1, ("Processing: [%2d] %s\n", depth, url));

	if ((up = url_open(url)) == NULL)
	    continue;

	/* search for more links from this one */
	gopher_enum(up, depth + 1);
	url_close(up);
    }

    finish_url();
    sigdie(0);
    /* NOTREACHED */
}